We are many instances of applications in a VM eviroment, now we have one tim attach every 3 instances.
We want to increase the a amount of instance of each app that they manage.
We are worried if every time he adds a new instance he must add a new tim, today we think that the tims are underutilized.
The thing is how many tims do we really need to monitor them applications, is there any way to get this estimate?
If we continue to grow in virtual instances of the application, where the tims should be located? so that a tim is not created per new instance deployed?
For the TIM capacity it is all about the volume of traffic data/packets needing to be processed subject to any web server filters you have applied.
Here is useful discussion on "TIM/MTP Administration Responsibilities" from Hallett_German which includes a link the docops CEM Capacity planning page. That link is actually broken so here is the correct version : CA CEM Capacity Planning - CA Application Performance Management - 10.5 - CA Technologies Documentation .
Also TIM SME JMertin might have more input.
I hope that helps
It would be really easier if the cemperfdb site would be reachable.
The thing with the TIM is that depending on
On a general rule, a regular TIM will handle 100Mbps of HTTPS traffic and up to 120Mbps HTTP traffic on a clean network tap based data feed.
Assuming the traffic is regular, and that none of the negative impact "feature" as listed before is active.
I have attached here a Percentile distribution of current performance data on TIMs I have.
Please remember, these are value extracted from real running TIMs, and that the throughput values are raw throughput (means, data thrown at the TIMs before the filter come into action).
If you would send me a data-set (GitHub - CA-APM/cem-healthcheck-scripts: Provides scripts to collect data for troubleshooting an APM 9.x or later instal… - please get me a CEMPERF performance data run), I could juxtapose it to the currently known TIM data I have to see how loaded/charged your current TIMs are.
Note, for the previous graph, take the following into account:
Hello Joerg and Lynn Thanks for your awnser i attach the link of file with the data of one of our tim.
We run the following option:
sample_data.tar.gz - Google Drive
Sorry for the time it takes guys, and we'll have to wait a little longer.
My local copy is a development version and has some bugs I was not able to get out yet,
and the transition prevents me to get the cemperfdb-server to upload the current data-set.
I'll come back to you as soon as I have access to both.
Its ok Joerg, thank you for you help
Thank you Joerg!
OK Guys, finally managed to get the app working again. When moving from mariadb 10.0 to 10.3 - the keyword "rows" could not be used anymore to name a temporary data extraction row. All of a sudden.
Anyway - here is the current percentile comparison:
The Red marked columns are where most data sets where found (Highest concentration), while the orange boxes show the data-set from yours.
In short - the throughput we have here is higher than average on VM (going from 33Mbps to max 64Mbps).
That would not be a problem, however the Dropped packets are showing there is an issue.
With this low throughput, the dropped packets usually come from broken packets as seen by the TIM Probe (You can read the implication here: The importance of clean traffic for APM CEM ) or from to the VM host not providing the appropriate resources to the TIM probe to do its job. When there is too much to do, and not enough CPU resources available (which the TIM/System does not see inside the VM), something has to give and if there is too much in the queue, the TIM will discard it.
Most of the rest is well within operational parameters.